Research And Comparison Of Short-term Load Forecasting Of Power System Based On Grey Theory And Neural Network

With the smart grid research and promotion of the construction, as well as the price ladder is about to commence in the national pilot, economic dispatch and safe operation of the proposed new requirements run on the power system, while the precision of load forecasting is accurate custom power systems the premise of the scheduling. Existing prediction methods have their own strengths, but none can guarantee that the results be served in any case. The key of short-term load forecasting of power system is collection of the advantages of existing technology to improve the prediction accuracy.There are many successful applications in load forecasting of power system based on grey theory and neural network. In order to improve the accuracy, combining with intelligent algorithm developed rapidly in recent years, such as genetic algorithm and particle swarm optimization, which has the characteristics of fast calculation and highlight global optimization ability, to optimize the forecasting model and parameters in this paper.Firstly, outlines the development status of the power system load forecasting, summarize the characteristics of the various forecasting methods and the power system short term load forecasting for power system security and stability of the significance. Analysis of the characteristics of the load data, then identificate and correct of abnormal data in the historical load data, and normalize the weather and holidays data. Secondly, respectively uses three methods of the grey theory, neural network combination forecasting and PSO-SVM to do modeling and forecasting, and compares the obtained prediction results. In neural network combination forecasting method, the RBFNN, GRNN and PNN are used for load forecasting respectively. Then, used genetic algorithm to optimize the combination weights dynamically. Finally, the three methods were developed a load forecasting software based on MATLAB. Calculation examples indicate that the forecasting results are stable and the precision meet the requirements.